Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Source. S. Fan, Strongly complete sets and a conjecture of Erdős, arXiv:2607.14071v5 (16 September 2026); Remark 4.2 on p. 20 (the second paragraph of Remark 4.1 of v4, p. 19, with the same content), with the definition (1.5) on p. 3, the definitions (1.8) and (1.9) and Corollary 1.2 on p. 4 and the discussion of Hegyvári's conjecture on p. 4. The artifacts are identified on the source card.

Read depth. Claims checked: the remark, the definitions and Corollary 1.2 were read clause by clause in the text layer of v5 and compared with v4; the remark's half-page argument was read through and not independently reviewed; Corollary 1.2 rests on Theorem 1.1, whose proof (Section 4) was not read. A preprint.

Statement

Definitions (pp. 3--4). For A⊆NA\subseteq\mathbb N, (1.5) (p. 3) is ∑a∈A∥aθ∥=∞\sum_{a\in A}\|a\theta\|=\infty for every θ∈T∖{0}\theta\in\mathbb T\setminus\{0\} (equivalently, the paper's spectrum H1(A)H_1(A) is {0}\{0\}). Mρ∗M_\rho^* (1.8) is the smallest positive integer MM with this property: whenever AA satisfies (1.5) and ∣A∩(ρk,ρk+1]∣≥M|A\cap(\rho^k,\rho^{k+1}]|\ge M for all large enough kk, AA is strongly complete. For α,β>0\alpha,\beta>0,

Aα,β={⌊2kα⌋,⌊2kβ⌋:k≥0}∖{0}(1.9);A_{\alpha,\beta}=\{\lfloor2^k\alpha\rfloor,\lfloor2^k\beta\rfloor:k\ge0\}\setminus\{0\} \qquad(1.9);

α∼β\alpha\sim\beta means α/β=2n\alpha/\beta=2^n for some n∈Zn\in\mathbb Z, and α\alpha is a dyadic rational when α∼n\alpha\sim n for some nonzero integer nn. Hegyvári's conjecture, as the paper reports it (p. 4): if α≁β\alpha\not\sim\beta and α,β\alpha,\beta are not both dyadic rationals, then Aα,βA_{\alpha,\beta} is complete. Corollary 1.2 (p. 4). If A⊆NA\subseteq\mathbb N satisfies (1.5) and has at least five elements in (2k,2k+1](2^k,2^{k+1}] for all large enough kk, then AA is strongly complete; that is, M2∗≤5M_2^*\le5. Remark 4.2 (p. 20). If M2∗M_2^* were 22, then Aα,βA_{\alpha,\beta} would be strongly complete, and so complete, for all α,β>0\alpha,\beta>0 with α≁β\alpha\not\sim\beta that are not both dyadic rationals; Hegyvári's conjecture would follow.

Proof pointer

Remark 4.2 (p. 20). Label the parameters so that α\alpha is not a dyadic rational, and write UK(x)={⌊2kx⌋:k≥K}U_K(x)=\{\lfloor2^kx\rfloor:k\ge K\}. Because α≁β\alpha\not\sim\beta, the rays U0(α)U_0(\alpha) and U0(β)U_0(\beta) meet in a finite set. Multiplying α\alpha and β\beta by suitable powers of 22 moves both into (1/2,1](1/2,1], where they differ (as α≁β\alpha\not\sim\beta) and α\alpha is still not a dyadic rational; this changes Aα,βA_{\alpha,\beta} by finitely many elements, which affects neither (1.5) nor strong completeness. Then, for every large enough k0k_0, Aα,βA_{\alpha,\beta} is the disjoint union of Uk0(α)U_{k_0}(\alpha), Uk0(β)U_{k_0}(\beta) and a finite set BB, and for each k≥k0k\ge k_0 the integers ⌊2k+1α⌋\lfloor2^{k+1}\alpha\rfloor and ⌊2k+1β⌋\lfloor2^{k+1}\beta\rfloor differ and both lie in (2k,2k+1](2^k,2^{k+1}], so each such interval holds at least two elements of Aα,βA_{\alpha,\beta}. Since α\alpha is not a dyadic rational, ⌊2k+1α⌋=2⌊2kα⌋+1\lfloor2^{k+1}\alpha\rfloor=2\lfloor2^k\alpha\rfloor+1 for infinitely many kk; for each such kk and each θ∈T∖{0}\theta\in\mathbb T\setminus\{0\} the triangle inequality gives ∥θ∥≤∥⌊2k+1α⌋θ∥+2∥⌊2kα⌋θ∥\|\theta\|\le\|\lfloor2^{k+1}\alpha\rfloor\theta\|+2\|\lfloor2^k\alpha\rfloor\theta\|, so the sum in (1.5) diverges. With two elements in every such interval and (1.5), M2∗=2M_2^*=2 would make Aα,βA_{\alpha,\beta} strongly complete. Read through; not reviewed.

Dependencies

Theorem 1.1 and Corollary 1.2 of the same paper for M2∗≤5M_2^*\le5; v5's Remark 4.1 for M2∗≥2M_2^*\ge2; Hegyvári's 1989 paper for the conjecture and its proved case (cited as the paper's [25]; not held).

Bears on

  • Problem 354: context. The remark's hypothesis M2∗=2M_2^*=2 is unproved (2≤M2∗≤52\le M_2^*\le5 is what the paper gives), so it resolves neither the irrational-ratio question, which the site-accepted Lean proof on the Conjectures.io card answers, nor the rational-ratio cases of Hegyvári's conjecture with neither number a dyadic rational (the paper reports that Hegyvári confirmed the case where exactly one is, p. 4), which remain open; the bounty site's review cites the paper (v4) as leaving "the relevant two-ray case unresolved".